Warning: The SGCCA algorithm did not converge

Analysis
1

Dear MixOmics Team:

While running the following function, I see the a warning populating many and many times:

test.keepX ← list(microbiome = c(5:9, seq(10, 30, 2)),
proteome = c(5:9, seq(10, 40, 3)))
tune.diablo2 ← tune.block.splsda(diablo.matrix, predictor, ncomp = ncomp,
test.keepX = test.keepX, design = design,
validation = ‘Mfold’, folds = 3, nrepeat = 20,
progressBar = T,
dist = “mahalanobis.dist”,measure = ‘BER’)
list.keepX2 ← tune.diablo2$choice.keepX

Warning: The SGCCA algorithm did not converge (occurred numerous times)

I think this might be caused by the small sample size I have (24 samples in total, with 15 in group 1 and 9 in group 2), but I am not exactly sure. I was wondering what this warning suggests, and how this influences the results? For example, this step is for finding the optimal number of features to include in each PC. So when the warning appears for an iteration (a combination of feature number), does it mean that the iteration fails and the feature number combination is not saved in
tune.diablo2$choice.keepX?

Thank you!

2

Hi @garyz,

Have a look at this related post. This warning might have appeared because some of the tested components are not informative (what did you input as ncomp?). If the tuning ran without errors (just warnings) you should have the results in tune.diablo2$choice.keepX, are you able to extract these and also plot the output of the tuning?

Cheers,
Eva